|
Critics Of Counterfactual Movement In Epidemiology Say “Pragmatic
Pluralism” Is Better Approach To Causal
Inference
“We wish to
forestall the emergence of a ‘hardline’ methodological school within
epidemiology, one which we feel would damage the discipline if it
became the dominant paradigm.” This is how authors Jan
Vandenbroucke, Alex Broadbent, and Neil Pearce
express their reason for writing a critique of the counterfactual
movement in an advance access article of the International Journal of
Epidemiology in January.
Non-trivial Damage
And
the damage the authors foresee is not trivial since the restricted
potential outcomes approach (RPOA) they criticize limits, in their
view, the questions that epidemiologists can ask, the designs they can
create, the evidence they consider acceptable for determining
causality, and consequently the evidence acceptable for scientific and
public health decision making.
Concerns
According to the authors, the counterfactual movement
is becoming dominant in theoretical epidemiology but it takes a too
narrow view of causality and fails to take into account the
contributions that can be made from diverse forms of evidence.
Interestingly, the authors differentiate “practical” causal inference
from a presumably more theoretical one which underlies the
counterfactual approach. They describe the historical practice of
epidemiology as “incredibly rich and successful” and believe that the
principles that are associated with the counterfactual or RPOA
approach “bear little resemblance” to the public health problem
solving that epidemiology has contributed to so usefully.
Formula 1 Cars
To make their point more colorfully, the authors employ
the metaphor of Formula 1 cars.
“Formula 1 cars may be the best in the idealized
environment of a racetrack, but to say that they are the ‘best cars’
would be misleading, since they are useless in almost every other
situation.”
Problems With RPOA
The criticisms of the RPOA approach for both the theory and practice
of causal inference in epidemiology are described in detail in the
paper. In summary, they are:
1. Restricting epidemiology to study feasible human
interventions misses the chance to study those which are not humanly
feasible such as heat waves or being born with two x chromosomes.
2. States such as “obesity” cannot be studied as causes.
3. Claiming that causation is well defined only when interventions are
well specified is problematic because the term “well-specified
intervention” itself is in need of definition.
4. RPOA ranks evidence in a way that ignores the context-dependence of
evidence.
5. RPOA does not make provisions for ruling out alternative
hypotheses.
6. It does not help in the use of different kinds of evidence to
arrive at one verdict about causality.
Pragmatic Pluralism
In place of RPOA, Vandenbroucke et al. argue for a
pragmatic pluralism which is less tied to a single concept of
causality. They state “…it is very plausible that we think about
causation in more than one way.”
They lay out a blueprint for teaching epidemiology and
what students should learn, namely:
1) that causal inference remains a judgment based on
intergration of diverse types of evidence,
2) to use diverse strategies to assess causality by
ruling out alternatives
3) about the elements of all types of epidemiological
study designs, inclusive of those types of design that do not match
the ideal counterfactual situation
4) to reflect critically on whether potential biases
matter ■
|